Skip to main content
Fig. 3 | BMC Cancer

Fig. 3

From: Bayesian copy number detection and association in large-scale studies

Fig. 3

Study site does not capture the major sources of technical variation. Hybridization intensities were available for four probes in a CNP region on chr4 spanning 9,370,866 bp - 9,410,140 bp (CNP_051). Restricting our analysis to high quality samples, we used the first principal component (PC1) as a one-dimensional summary of the 4 x 6,026 matrix of log2R ratios. The density of the PC1 summaries marginally (black) and stratified by study site (gray) are bimodal, suggesting a copy number polymorphism a. However, stratification of the PC1 summaries by grouping chemistry plates with similar eCDFs reveals an obvious batch effect (b). For example, chemistry plates in group E comprised of 786 samples originating from all nine study sites has a markedly different distribution than the 567 samples processed on group C chemistry plates

Back to article page